DUNCAN PRITCHARD

Reforming Reformed Epistemology*

0. Introduction
There has been a renaissance of interest in the epistemology of religious
belief over the last twenty years which has been largely inspired by the
work conducted by Alvin Plantinga, William Alston, Nicholas Wolterstorff
and others on the so-called reformed defence of the rationality of religious belief. The starting-point for this reformed conception of religious
epistemology is a rejection of the supposedly evidentialist assumptions
which drive standard sceptical arguments regarding religious belief.1 I
think that this general negative claim is, in its bare essentials, correct, although canvassing support for an argument for this contention will not be
my primary concern here. Instead, I will be outlining one way in which the
reformed epistemological stance can be modified to make it resistant to a
certain sort of attack. I suggest that the manner in which the reformed conception of the epistemology of religious belief is often motivated with respect to a supposed analogy with perceptual belief has tended to overemphasize certain features of religious belief, and that recognising this fact
enables one to offer a more fine-grained account of the epistemic status of
religious belief. In particular, I argue that making this point clear draws out
the sense in which reformed epistemology, properly understood, should be
allied to a specific form of virtue epistemology.
In 1, I give an overview of the standard sceptical attack on the rationality of religious belief and briefly outline the sense in which such
scepticism is held to rest upon an evidentialist thesis. In 2, I elucidate the
manner in which the theses of evidentialism, classical foundationalism and
epistemological internalism are intertwined within the debate between the
reformed epistemologist and the sceptic about religious belief. In 3-4, I
explain how reformed epistemology is typically motivated with respect to a
*

A slightly different version of this article was published (with the same title) in International Philosophical Quarterly 43, 2003: 43-66. I am grateful to the editors of
that journal and to the publishers, the Philosophy Documentation Center, for permission to reprint this article here.
1
For an excellent overview of the debate, see Plantinga and Wolterstorff 1983.

176
parity argument concerning certain relevant analogies between religious
and perceptual belief, and I go on to consider some problems that this parity argument faces. In 5, I consider Keith DeRoses intriguing suggestion
that one can resolve some of these problems by reconfiguring reformed
epistemology in terms of the sort of foundherentist epistemological
model advocated by Susan Haack. I argue that this proposal should be endorsed provided that it is understood at a very general schematic level. So
construed, however, it fails to offer us the fuller account that we are after.
Accordingly, in 6, I argue that the best way of adding content to this
foundherentist structural model of religious epistemology is by conceiving
of reformed epistemology in terms of a specific type of virtue epistemology. Finally, in 7, I offer some concluding remarks.
1. Scepticism about Religious Belief
Perhaps the most common form of scepticism about religious belief has an
ontological rather than (at least directly) an epistemological form. That is,
the focus of the attack is not directly on the epistemic status of the belief in
question but rather concerns the ontology that is thought to underlie that
belief. One might naturally think that any dispute over ontology would always spill-over into a dispute about epistemic status. If I think that you are
wrong to commit yourself to a certain ontology then do I not thereby regard your beliefs in such an ontology as lacking the requisite epistemic
status? Interestingly, the answer to this question is, despite first appearances, no. Think, for example, of two eminent scientists who, whilst
sharing many beliefs, have differing beliefs regarding the ontological
status of a certain entity that is sometimes postulated by scientists working within a particular area of scientific inquiry, the one thinking that it exists and the other thinking that it doesnt. Although each is committed to
regarding the other as being in error in some way, both of them could recognise that the others belief had some substantial degree of positive epistemic status. Why, then, is it the case that when it comes to disputes between religious believers and their detractors, the latter tends to formulate
the ontological challenge such that it has direct, and devastating, epistemological ramifications for the epistemic status of the religious beliefs in
question? The standard answer is that the difference between the scientists
in our imagined example and the religious believer is that the scientists can
in principle offer appropriate evidence to support their belief whereas the
religious believer cannot.

177
Consider the contrast between the two cases. In the case of the scientists, each of them could, conceivably, offer the other evidence which they
both accept as being good evidence for the belief in question (even though,
of course, they are both obliged, as long as they maintain their positions, to
disagree about the extent of the relative evidential support in each case).
The situation facing the religious believer is very different. The only evidence that he could propose which might potentially be adequate for the
task will be contentious in this context. He might base his religious beliefs
upon certain sorts of religious experiences, revelation for example, but this
sort of evidence will only be apt in this context provided that one is already
willing to grant the ontology that is presupposed by counting these experiences as religious in the relevant sense.2 Indeed, the only sort of evidence that would seem to be appropriate for this purpose would be an a
priori argument of some description, or at least some sort of incorrigible
empirical evidence, and few people would now hold that there are grounds
of this sort available to license religious belief. There is thus a sense in
which the religious believer is unable to adduce non-question-begging evidence in support of his religious belief.3
In both cases the dispute is being characterised in terms of an evidentialist requirement to the effect that one should proportion ones belief to
the evidence available to support that belief, but the thought is that whereas
certain sorts of beliefs can, in principle, meet this requirement, there is an a
2

At least in the standard case. Of course, there may be certain sorts of events which
everyone might be willing to recognise as providing corrigible empirical evidence in
support of religious belief that is non-question-begging (the sky opens up across the
world and a being comes down proclaiming certain verses from a key religious text,
and so on). But even if one is willing to grant that such cases might be able to offer the
requisite non-question-begging evidence, this concession will offer little comfort to the
apologist for religious belief. After all, the epistemic status of ones religious belief
would then just be hostage to the occurrence of such events and, in the meantime,
completely lacking.
3
In this respect scepticism about religious belief mirrors more general forms of scepticism. Consider external world scepticism, for example, where the claim is that one is
unable to adduce the kind of non-question-begging evidence required which would
support the belief without already assuming (as do most empirical beliefs) the existence of an external world in the first place. Accordingly, any anti-evidentialist strategy as regards religious belief will have application to scepticism in general. For more
discussion of this question-begging aspect of radical sceptical arguments, see Wright
1985; 2000 and Pritchard 2002a. For a discussion of the general relationship between
scepticism about religious belief and radical scepticism, see Pritchard 2000. For a survey of recent work on radical scepticism, see Pritchard 2002b.

178
priori difficulty with religious belief which means that it can never adequately fulfil it. We shall explore the thesis of evidentialism more fully below, but it is worth noting that, at first glance at any rate, the evidentialist
requirement seems an entirely uncontentious hurdle for religious belief to
clear.
2. Evidentialism, Classical foundationalism and epistemological
internalism
Historically, one common response to this sort of sceptical argument has
been to try to supply the very sort of a priori grounds that are being demanded. In contrast, another type of response has been to embrace the
sceptical conclusion of that argument by claiming that religious belief is
not the sort of belief that requires epistemic support.4 What both of these
approaches to scepticism about religious belief share is an implicit acceptance of the nature of the sceptical argument such that any response to that
argument must consist in either some sort of rapprochement with the conclusion or a rejection of the key premise that religious beliefs lack the evidence in question. In contrast, the reformed epistemological stance under
discussion here adopts the completely different approach of calling into
question the very evidentialist doctrine that drives the sceptical argument.5
In many of the key accounts of reformed epistemology, the locus for
evidentialism has been W. K. Cliffords provocative claim that [I]t is
wrong always, everywhere, and for anyone, to believe anything upon insufficient evidence6. There are, however, eminent proponents of the evidentialist thesis in more recent epistemological debate, so it is perhaps better to focus on one of these contemporary accounts. For example, in two
recent papers (one of them co-authored with Earl Conee), Richard Feldman
has offered a spirited defence of the evidentialist position. He defines the
thesis as follows:
For any person S, time t, and proposition p, if S has any doxastic atti4

This style of response has, of course, found a great deal of support amongst those
guided by what they consider to be certain insights in Wittgensteins later remarks on
religious belief, especially as they appear in Wittgenstein 1966. See, for example,
Nielsen 1967 and Phillips 1976.
5
Note that these three alternative defences of religious belief are not all in competition. One could, for instance, argue that there are a priori grounds available in support
of religious belief whilst also contending that evidentialism is an erroneous doctrine.
6
Clifford 1879: 186.

179
tude at all toward p at t and Ss evidence at t supports p, then S epistemically ought to have the attitude towards p supported by Ss evidence at t.7
Although this formulation may meet some of the objections that are most
naturally levelled at the much less precise rendering of the doctrine due to
Clifford, it still faces several seemingly insuperable difficulties.
To begin with, we need to look a little closer at how we are to understand the notion of evidence in play here. Presumably, we cannot just understand an agents evidence to be whatever grounds that are available to
the agent in support of his belief regardless of whether or not he attends to
(or even can attend to) that evidence. There are many reasons for this, but
the most pressing is that we want the evidence to play the required supporting role, and unless that evidence is attended to then it need not perform
this function. Consider, for example, the case of a man who has excellent
evidence in support of his belief that a certain person is guilty, but who
does not attend to this evidence when forming the belief. Instead, he forms
his belief as a result of a bigoted ill-feeling towards the person in question.
Clearly, this would not be a case of someone meeting the kind of epistemic
requirement that is meant to be imposed by evidentialism because this persons belief would not enjoy any substantial measure of positive epistemic
status, despite the existence of evidence in support of that belief that is
available to the subject. As Feldmans characterisation of evidentialism
stands, however, it makes no distinction between an agent who believes
what he epistemically ought to believe because of bigoted ill-feeling and
an agent who believes what he epistemically ought to believe because he is
sensitive to the evidence that is available to him in support of that belief.8
It appears, then, that we must restrict the evidentialist thesis so that it
is understood as requiring that an agent should proportion his belief to the
evidence that he has for that belief and which he is presently attending to.9
Even under this construal, however, problems remain. For unless we offer
7

Feldman 2000: 679; cf. Feldman and Earl Conee 1985.

This example is adapted from one used by DeRose 2000, 697-8, in his astute response to Feldman 2000. For some of the other, to my mind, decisive criticisms that
can be levelled against evidentialism, one should consult DeRoses paper.
9
Of course, a more detailed characterisation of the evidentialist account would involve an elucidation of this notion of attending to. Since I am not defending the evidentialist position, however, and since I do not think that anything consequential hangs
upon our understanding of this notion, I will not be undertaking a detailed discussion
of it here.
8

180
some restriction on what can count as evidence in this respect then it is far
from clear that evidentialism, so understood, could plausibly drive a sceptical argument regarding religious belief. After all, as it stands, this characterisation of evidentialism would seem to permit one to adduce such evidence as revelation in support of ones religious belief and, as noted above,
the religious believer has (by his own lights) plenty of evidence of this
sort. How, then, do we form the connection between a general evidentialist
thesis and scepticism about religious belief?
The standard answer to this question is to argue, whether explicitly
or implicitly, that it is not evidentialism alone that is the motivation behind
scepticism about religious belief, but rather the combination of evidentialism and what Plantinga refers to as classical foundationalism. He characterises the classical foundationalist account in terms of a certain conception
of the criteria that a belief needs to meet if it is to be properly basic:
For any proposition A and person S, A is properly basic for S if and
only if A is incorrigible for S or self-evident to S.10 11
It is certainly true that, historically, classical foundationalism and evidentialism have been closely intertwined. Although there is some debate about
the historical roots of the evidentialist requirement on religious belief, it is
certainly present in the work of Locke, and, significantly, Locke combines
such a thesis with a version of classical foundationalism that demands that
the foundational propositions be certain.12
Locke famously wrote in the Essay Concerning Human Understand13
ing that reason must be our last judge and guide in everything, and, accordingly, he maintained that religious beliefs should be put before the tribunal of reason just like any other. This line of thinking did not lead to
scepticism about religious belief as far as Locke was concerned, however,
because he held that the necessary evidential grounds were available in
support of religious belief. The conclusion was thus only that we must dis10

Plantinga 1981: 49
In later works, such as Plantinga 1983, Plantinga extends the class of properly basic
propositions to include those propositions that are also evident to the senses.
12
More specifically, as Wolterstorff 1991, 81, puts it:
Lockes official view is that the only things we know immediately are those necessary truths that are self-evident for us, whereas his unofficial view is that we can know
immediately whatever we are certain of without inference. In either case, knowledge is
grounded in certitude.
13
Locke 1979 IV, xix: 14
11

181
tinguish between those believers whose religious beliefs were grounded in
reason and those believers (whom Locke called the enthusiasts) whose
religious beliefs were grounded only in revelation. He describes those who
fall into this latter camp as follows, arguing that if they regard what they
believe as being true solely
[...] because it is a revelation, and have no other reason for its being a revelation
but because they are fully persuaded, without any other reason, that it is true,
they believe it to be a revelation only because they strongly believe it to be a
revelation; which is a very unsafe ground to proceed on, either in our tenets or
actions.14

What we find with Locke is the claim that evidence must be available to
support the religious belief in question and that the only evidence suited to
the purpose, given his prior commitment to (one form of) classical foundationalism, will be that which is derived from propositions which are certain. Although this line of thought is, of course, distinct from scepticism,
one need only remove the conviction that the relevant certain foundational
propositions are available for the sceptical challenge to emerge.15
Likewise, in contemporary discussion, the thought is that the legitimation of religious belief requires not just evidence but evidence of a certain kind: foundational evidence of an a priori or incorrigible variety. So
construed, the connection between evidentialism and scepticism about religious belief becomes transparent.
Nevertheless, one should be careful about identifying evidentialism
too closely with classical foundationalism because this can tend to obscure
the underlying nature of the dialectic in play here. In particular, too close
an identification can lead to the impression that merely denying classical
foundationalism would suffice to meet this sceptical challenge. This conception of the debate is certainly wrong, however, because one could state
the evidentialist challenge without making any mention of classical foundationalism.
In order to see this point, it is worthwhile returning to the contrast
drawn earlier between the two scientists who were arguing about the ontological status of a certain entity and the debate between the religious be14

Locke 1979 IV, xix: 11

There are, of course, subtleties to Lockes view in this respect that this brief overview cannot do justice to. Nevertheless, this short account should suffice for our purposes here. For an excellent discussion of Lockes view in this regard that is sympathetic to the line taken here, see Wolterstorff 1996.
15

182
liever and the sceptic about religious belief. Significantly, in both cases
one can characterise these debates in terms of an evidentialism that does
not incorporate a commitment to classical foundationalism. Take the scientific case first, which was meant to be a situation in which one could, in
contrast to the religious case, respond to the evidentialist challenge with
the requisite evidence. It seems perfectly acceptable in such a scenario to
regard one of the disputants as possessing the greater degree of evidential
support for his belief, and thus as being able to convince the other scientist
that he should weaken, if not completely change, his opposing belief.
Moreover, there need be no assumption in play here that the evidence adduced should be a priori or incorrigible. Indeed, one would expect it to be
ordinary corrigible empirical evidence. Of course, one might further demand that, at some point, this chain of support must lead to incorrigible or
a priori evidence, but there need be no mention of this claim directly and it
is far from obvious that such a further move is entailed by the mere evidentialist thesis alone.
Similarly, one can characterise the debate about religious belief
without mention of classical foundationalism. For on at least one plausible
construal of evidentialism, which demands not just evidence but nonquestion-begging evidence, then it will be true that the religious believer
will be unable to offer the requisite grounds. Moreover, we can put this
point in terms of evidentialism alone without making any direct recourse to
classical foundationalism because there is no obvious sense in which this
non-question-begging demand must entail a demand for a priori or incorrigible evidence.
Of course, as we saw above, the thought that the reformed epistemologist has is that this non-question-begging form of evidentialism is
simply the result of conjoining a more neutral form of evidentialism with
the classical foundationalist thesis, and thus that classical foundationalism
is implicit within this purely evidentialist account of religious scepticism
after all. The reformed epistemologist will typically claim, for example,
that when it comes to religious belief the only way to offer such nonquestion-begging evidence is to adduce a priori or incorrigible grounds,
and thus that this variety of evidentialism presupposes classical foundationalism. On this understanding of the debate, it is merely a matter of taste
whether one characterises the sceptical argument in terms of a beefed-up
evidentialism or in terms of a neutral rendering of the evidentialist doctrine
which one then supplements with classical foundationalism.
Such a conception of the debate can tend to mislead, however, since

183
what ultimately underlies the move from the neutral form of evidentialism
to the non-question-begging version is not classical foundationalism at
all but rather epistemological internalism.
Epistemological internalism is, roughly, the thesis that positive epistemic status demands reflective access on the part of the subject to those
facts that determine that epistemic status.16 In the case of evidentialism, for
example, an internalist variant of this thesis would not just demand the
availability of evidence to the subject, but also that the subject is actually
in a position to not only reflectively access that evidence but also reflectively determine that it is evidence. This is the kind of demand that is being
made in the more restricted version of evidentialism just considered where
the agent has to actually vouch for the evidence as being evidence that is
suited to the purpose (rather than question-begging evidence).
Evidentialism and epistemological internalism tend to go together
because any natural rendering of the evidentialist thesis invokes the internalist doctrine.17 Nevertheless, it is best to keep the two theses apart as
much as possible to bring out the role that epistemological internalism
plays in the sceptical argument regarding religious belief. In particular, by
being clear about the role played by epistemological internalism in this respect, we avoid the temptation of thinking that a simple denial of classical
foundationalism would suffice to meet the sceptical problem about religious belief. For even if one granted that corrigible empirical nonquestion-begging evidence could support legitimate belief (as we granted
above in the case of the two scientists), it would still remain that we have a
sceptical problem regarding religious belief because of the inherently question-begging nature of the evidence involved.
Note that this is not to deny that the role that classical foundationalism plays in the sceptical argument about religious belief is closely intertwined with that played by epistemological internalism. After all, the most
obvious way of meeting this non-question-begging demand is by adduc16

Though an extremely rough characterisation of this doctrine, it should suffice for

our purposes here. The locus classicus for epistemologically internalist approaches is,
of course, Chisholm 1989. For more on the epistemic externalism/internalism contrast,
see the papers contained in the excellent anthology on this subject edited by Kornblith
2001.
17
One reason for this is that evidentialism tends to be closely associated with a certain
deontological thesis regarding justification that has explicitly internalist overtones. Indeed, Feldman e.g., 2000 is a good example of someone who holds both an epistemologically internalist thesis characterised along deontic lines and, for related reasons,
also endorses a form of evidentialism.

184
ing a priori or incorrigible empirical evidence in the manner that classical
foundationalism demands. But this merely reflects the fact that classical
foundationalism tends to be a natural consequence of epistemological internalism, and thus that it is epistemological internalism that is the underlying force behind scepticism about religious belief.
That scepticism about religious belief should ultimately find its
source in epistemological internalism should not surprise us, since, as is
well-known, epistemological internalism drives a number of other sceptical
arguments as well.18 Moreover, I think we gain a better understanding of
the core elements of the reformed epistemology response to scepticism
about religious belief by understanding it as, primarily, an epistemologically externalist response to such scepticism, rather than as an antievidentialist or anti-classical-foundationalist thesis. Accordingly, henceforth when I refer to evidentialism it will be the explicitly internalist variant of this thesis that I have in mind.
3. The Parity argument
One of the great benefits of being clear about the nature of the evidentialist
demand as regards religious belief is that it highlights just how austere that
demand is. The reformed epistemologist can use this observation to his advantage. In particular, reformed epistemologists have pointed out that such
a requirement on positive epistemic status would seem to rule-out a great
deal of belief as being epistemically lacking. The standard example employed here is that of perception since, if we know anything much about
the world at all, then it would seem that we must have some perceptual
knowledge. Consider how perceptual knowledge fares with respect to the
evidentialist thesis, however. Is it really true that we are able to offer evidence that we can vouch for which is sufficient to support the level of conviction present in our standard perceptual beliefs?
In order to make the parallel between the religious case and the perceptual case more direct, consider someone who is sceptical about perceptual belief in general on the grounds that he is sceptical about the external
world ontology that it commits one to. How would one convince such a
sceptic? Clearly, as with the religious case, it would not do to cite empiri18

Plantinga certainly would not be surprised by this observation, of course, because it

informs a great deal of his discussion of warrant in Plantinga 1993a. Nevertheless, as I
will be arguing below, Plantinga has not adequately recognised the nuanced manner in
which the externalism/internalism distinction impacts on this debate.

185
cal grounds that are gained by perception since this would be questionbegging. This evidence is only apt for the purpose provided that the beliefs
at issue are genuine perceptual beliefs (rather than merely beliefs about
how things seem), and insofar as they are assumed to be genuine perceptual beliefs then the ontology in question is being illicitly taken for granted.
As with the religious case, then, the only way around this requirement
would appear to be to offer either a priori or incorrigible support for our
perceptual beliefs, and such support is not obviously forthcoming. Allowing evidentialism in the case of religious belief would thus seem to license,
by parity of reasoning, radical scepticism as regards perceptual belief. Provided that we grant that if we know anything much at all then we must
have some perceptual knowledge, we thus have independent grounds to be
sceptical about the evidentialist requirement itself.19
Reformed epistemologists have been keen to exploit this parallel between religious belief and perceptual belief as a means of putting religious
belief on the same sort of secure footing that is typically granted to perceptual belief. This line of reasoning has been characterised by Alston and
others as a parity argument. Alston argues that the point of such an argument is to show that what he calls Christian Practice
[...] has basically the same epistemic status as [Perceptual Practice] and that noone who subscribes to the latter is in any position to cavil at the former.20,21

In order to achieve this end, the possibility which Alston explores is that
[...] religious experience is basic to religious belief in somewhat the way in
which sense experience is basic to our beliefs about the physical world. In both
cases [...] we form certain beliefs about the subject matter (God, the physical environment) on the basis of experience [...].22

19

Of course, the religious sceptic could at this point retreat into a general scepticism
that was not solely confined to religious belief. The trouble with this manuvre, however, is that it only serves to undermine the original sceptical argument about religious
belief. After all, that argument contended that there was something peculiarly defective about religious belief, not that belief in general was problematic. Indeed, if belief
in general is problematic then the religious believer is no worse off than the agnostic
or the atheist. Newman is sensitive to this point. See Newman 1844; 1985.
20
Alston 1982: 12
21
The original quotation has the former and latter in the reverse order but, as
DeRose 1999a notes, this is surely a mistake on Alstons part.
22
Alston 1986: 2

186
In particular, reformed epistemologists tend to concentrate on two parallels
between religious and perceptual belief.
The first parallel, which we have already noted, is that both religious
beliefs and perceptual beliefs are prone to an evidentialism-based sceptical
argument. Insofar as one regards perceptual belief as legitimate, then, one
is obliged to regard religious belief as being, at least prima facie, legitimate as well. That is, if evidentialism is inapplicable as regards perceptual
belief, then it cannot simply be assumed to play a role as regards religious
belief. This is the negative element of the parity argument since it merely
contends that there are prima facie grounds for thinking that religious belief is no worse off, epistemically speaking, than perceptual belief.23
The second parallel is more positive in that it draws upon relevant
similarities between the nature of perceptual experience and the nature of
religious experience which would appear to license the same sort of nonevidentialist epistemology that is often applied in the perceptual case to the
religious case.24 More specifically, the thought is that religious belief can
sometimes enjoy the very sort of directness that is often found in perceptual belief and which, in the perceptual case, licences the adoption of a
non-evidentialist epistemology. As Laurence Bonjour25 has put it, ordinary
perceptual beliefs tend to arise spontaneously out of certain perceptual
experiences, and the same might be said regarding how certain religious
beliefs arise in response to particular religious experiences. By exploiting
this positive analogy between religious and perceptual belief, reformed
epistemologists, such as Alston and Plantinga, have argued that the kind of
non-evidentialist epistemological model that is applied in the perceptual

23

Another negative claim that Plantinga 1983 makes in this respect is that classical
foundationalism is self-refuting, because an agents belief in the classical foundationalist doctrine will not itself be grounded in either incorrigible or a priori evidence.
24
What is meant here by religious experience? The best account that I know of is
due to Alston 1986 who argues that it is those experiences that give rise to the Mbeliefs (or manifestation beliefs) that are the focus for a great deal of his discussion
on this matter. He writes that religious experience concerns:
[...] experiences that would naturally lead the experiencer to formulate that he has
experienced something about Gods current relation to himself; that God said to
him, that God was enlightening him, comforting him, guiding him, sustaining him in
being, or just being present to him. Alston 1986: 6
One of the advantages of this characterisation is that the beliefs that it gives rise to will
tend to have a clear propositional content.
25
BonJour 1985, passim.

187
case ought to be just as apt in the religious case.26
In particular, the failure of evidentialism as regards perceptual belief
points towards the need for a form of externalist epistemology that characterises perceptual knowledge in terms of an appropriate reliable relationship between ones beliefs and the physical environment which those beliefs are supposed to track. Such an account is externalist in the sense that
it does not require that the agent who has epistemically supported perceptual beliefs should be able to reflectively recover the grounds that support
such beliefhis evidencein the way that evidentialism demands. This
kind of approach to perceptual belief takes its cue from a certain construal
of the writings of Thomas Reid, and in effect maintains that perceptual belief enjoys a kind of default epistemic support such that, in the absence
of countervailing evidence and provided that the belief-forming mechanisms in question are as a matter of fact reliable, an agents perceptual belief can enjoy sufficient positive epistemic status even when the agent is
unable to adduce adequate evidential grounds to support that belief.
Prima facie, an externalist epistemology understood along these lines
ought to be equally applicable to religious belief. Just so long as the beliefforming mechanisms in question (however they are to be characterised) are
appropriately reliable, and just so long as there is no countervailing evidence to take into account which would imply that those mechanisms are
not functioning adequately, then such beliefs seem to enjoy a default positive epistemic status. Of course, anyone who does not share the religious
belief in question will be sceptical about the putative reliability of the belief-forming mechanisms just as they will be sceptical about the ontology
presupposed by such beliefs, but this situation is no different in relevantly
similar cases, such as with perceptual belief. Saying that perceptual belief
gives us knowledge so long as it is indeed reliable in the appropriate way
will not persuade someonethe radical sceptic, saywho doesnt already
accept the ontology that is being presupposed here. Nevertheless, such a
manuvre would support the claim that perceptual knowledge is at least
conditionally possible given that certain conditions actually do obtain.
Similarly, the religious believer can respond to an evidentialism-based
26

See Alston 1986; 1991, and Plantinga 1993a; 2000. As I discuss below with respect
to Alston, however, different reformed epistemologists have differing conceptions of
what can be concluded from this positive element of the parity argument. The reader
should also note that although it is common practice to refer to both Alston and Plantinga as reformed epistemologists, both of these writers have rejected this categorisation.

188
scepticism about religious belief by arguing that religious knowledge is indeed possible after all, just so long as certain factual conditions obtain.
In effect, what the religious believer is doing here is distinguishing
the epistemological question of whether it is ever possible for ones religious beliefs to enjoy a sufficient degree of positive epistemic status from
the metaphysical question of whether there really does exist the kind of ontology to which the religious believer is committed. Though a negative answer to the metaphysical question would, of course, prejudice the possibility of a positive answer in the epistemological case, in the absence of such
a negative metaphysical conclusion the epistemological question can be
met. The burden is thus placed back on the sceptic about religious belief to
offer the relevant negative metaphysical argument, and it is hard to see
how such an argument could be supported. Moreover, since the epistemological model in question is independently plausible in the case of perceptual belief, and since there are relevant similarities between the perceptual
and the religious case, hence it seems that the defender of religious belief
has made significant headway against the sceptic. The legitimacy of perceptual belief is rescued from the threat posed by an evidentialism-based
scepticism and thus, by parity of reasoning, so is religious belief.27
4. Problems with the Parity Argument
I think that this general externalist defence of religious belief on these
grounds of parity is persuasive. As so often in philosophy, however, the
devil is in the detail, since the problems arise once one opts for a specific
externalist construal of the epistemology of religious belief. In particular, I
want to focus upon the most common account that runs along these lines
that is due to Plantinga. His claim is that we should treat religious belief as
being properly basic in just the same way that perceptual belief is treated
as properly basic on the externalist account just considered. This claim of
27

For two interesting discussions, and overviews, of the kind of parity arguments employed by reformed epistemologists, see Penelhum 1986; cf. Penelhum 1983; 2000,
and McLeod 1993. In the former work, Penelhum also points out that Plantingas early
characterisations of the reformed epistemological position e.g. Plantinga 1983 were
misleading since they tended to imply that grounds were being offered to support religious belief over other sorts of non-religious belief. This is not the case, however,
since all that is achieved is a kind of advantageous impasse with the religious sceptic
(though this is achievement enough). I think that in later work Plantinga is more explicit about this point, though see DeRose 1999b for more discussion of this worry in
the light of Plantingas more recent work, in particular, Plantinga 2000.

189
proper basicality comes down to the contention that, although one might
have evidence in support of ones religious or perceptual beliefs, such beliefs (at least in favourable cases) do not stand in need of an evidential
grounding in order to be properly held. Provided that certain external conditions are met, a perceptual or religious belief can enjoy an immediate
warrant that arises directly out of the agent having an experience of a certain sort, rather than being a transferred warrant, or partly transferred warrant, which is dependent upon an evidential grounding.28
This line does more than merely rescue religious belief from an evidentialism-based scepticism, however, it goes further to actually allow, as
with perceptual belief, that evidence need play no substantive warranting
role when it comes to certain religious beliefs. It is this claim that religious
beliefs can be properly basic in this way that I think is questionable.
The problem with this construal of the epistemology of the religious
belief is that it overstates, in relevant respects, the parallels between religious experience and perceptual experience. In particular, there is a worry
about the putatively analogous spontaneity of religious and perceptual
beliefs. It was noted above that religious beliefs, like perceptual beliefs,
can sometimes seem to have the same sort of directness that one might
find in the perceptual case, as if one is directly responding to a religious
being in the way that one directly responds to objects in the physical world
via perception. It was this spontaneity of perceptual belief that made it apt
for a radically non-evidentialist construal since evidence seemed to play no
essential warranting role as regards standard perceptual belief. The problem, however, is that whereas this sort of directness is the norm in the
perceptual case, it is more naturally thought of as the exception to the norm
in the religious case. Indeed, whereas perceptual beliefs seem to be, in the
main, forced upon us, religious beliefs often seem to be formed in a far
less immediate and compelling fashion. As Keith DeRose29 has put it,
normal religious belief is rarely understood in terms of being zapped by
a divinity, as Plantinga seems to understand it; instead, the more common
way of conceiving of such belief is in terms of being nudged or invited
towards a certain sort of doxastic commitment.
In general, there does seem to be a certain voluntary element in re-

28

Following Plantinga 1993a I will understand warrant to be that epistemic notion

that is sufficient, with true belief, for knowledge.
29
DeRose 1999a.

190
ligious belief that is absent in most forms of perceptual belief.30 Relatedly,
whereas an agent might gain a warrant for his perceptual belief without engaging his reflective capacities at all, religious beliefs seem to directly implicate such capacities. For example, whereas we are happy to attribute
perceptual knowledge to small children, we are unwilling, I think (except
in rare cases), to ascribe anything but the most basic of religious knowledge to persons who lack reasonably developed reflective capacitates. This
hints towards the fact that well-formed religious beliefs seem to demand
more of the subject than well-formed perceptual beliefs.
Of course, for the parity argument to go through all that is needed is
that some religious beliefs are basic in the same way that some perceptual
beliefs are; it is not essential that religious belief be in general analogous to
perceptual belief. Indeed, Plantingas own discussion of the role that defeaters can play regarding basic religious beliefs would appear to indicate
that the religious beliefs held by most reasonably sophisticated religious
believers are, for the most part at least, non-basic.31 Nevertheless, the concern about the disanalogies between religious belief and perceptual belief
is important because it directs us to look again at the idea that, on grounds
of parity, we should consider religious beliefs to be basic. After all, if we
concede that properly formed religious belief may be constrained by more
imposing demands than properly formed perceptual belief, then it becomes
far from clear that it follows from the fact that certain perceptual beliefs
are basic that any (or hardly any) religious beliefs are basic as well.
It is open to reformed epistemologists to query these putative intuitions, of course, or at least question the epistemological ramifications that
they are meant to hold. Rather than engaging in such dialectical warfare,
however, a better approach might first be to see whether the general re30

Note that by these remarks I am not committing myself to some sort of doxastic
voluntarism thesis. Rather, I am merely highlighting the fact that, insofar as we have
control over our beliefs at all, then it is part of our ordinary conception of our doxastic
capacities that we have more control over the formation of the standard religious belief
than we do regarding the formation of the standard perceptual belief. Moreover, it may
seem as if I am also endorsing some sort of general thesis to the effect that belief control is a direct result of the belief being formed in a non-spontaneous way. This is
not the case. My remarks here are solely confined to the cases of religious and perceptual belief.
31
Though not necessarily. A basic belief that has been subject to a defeater which, in
turn, has been defeated by a second defeater, will return to being properly basic. For
an excellent discussion and overview of the issue of the relationship between basicality and defeaters, see Miller 2004.

191
formed epistemology framework couldnt simply be modified, in nonessential respects, to enable it to accommodate these disanalogies. For if it
can then the dialectical warfare is unnecessary and the claim that there are
these disanalogies has constructive, rather than destructive, consequences
for the view.
One motivation for pursuing the irenic goal of integrating this concern into the reformed epistemology thesis rather than attempting to argue
it down, is that the intuitions that drive this concern find expression in
the work of one of the most prominent reformed epistemologists, Alston.
Not only has Alston noted these disparities between perceptual and religious experience but, as a result, he has advocated a far more cautious nonevidentialist epistemology as regards religious belief than one finds in
Plantinga. In particular, he argues that religious experience alone cannot
suffice to warrant ones religious beliefs, contending that further transferable epistemic support must be sought from other sources, such as salient
historical evidence and other relevant evidence that is socially transmitted
from other members of the agents religious community.32 That Alston is
willing to even consider such a move offers prima facie grounds for thinking that perhaps the basicality thesis does not play quite such a central role
in the reformed epistemology stance than is often thought.
I think that this is right, and that, properly understood, the driving
motivation for reformed epistemologyand which enables it to evade
scepticism about religious beliefis its commitment to epistemological
externalism. This point is in keeping with a claim made earlierthat, at
root, it is internalism rather than evidentialism that drives scepticism about
religious belief. Accordingly, the rejection of evidentialism contained
within reformed epistemology is not nearly so important as the move away
from epistemological internalism and towards epistemological externalism.
As a result, I will be arguing that provided this externalist element of the
view is retained then one can forge a version of the thesis that can accommodate these disanalogies. Note that in what follows I will allow that there
may be some religious beliefs that are basic in the appropriate sense. What
I will be claiming is that the reformed epistemological stance is not hostage to the existence of these beliefs since, even if no religious belief ever,
in fact, meets the basicality rubric, the general elements of the view can
still be regarded as secure and capable of evading the sceptical attack.33
32

See, in particular, Alston 1982; 1986.

DeRose 1999a suggests that the reason why Plantinga is not as sensitive to the disanalogies between religious and perceptual belief is that he has a very specific model

33

192
5. DeRoses foundherentist proposal
Before we consider how this modified reformed view is to work, it is
worthwhile looking at a suggestion made by DeRose that runs along similar lines. DeRose (1999a) argues that we can respond to these disanalogies
between perceptual and religious belief by understanding religious belief in
terms of the sort of foundherentist model advocated by Susan Haack
(1993; 1995). One effect of adopting this line of response is that it leads to
the rejection of the kind of foundationalism that is implicit within Plantingas version of reformed epistemology as it currently stands. In particular, the foundherentist idea as it applies to Plantingas account is precisely
not to treat religious beliefs as properly basic at all, except perhaps in rare
cases. This may not be quite such a dramatic change as it might at first appear, however, for two reasons. First, because foundherentism does incorporate some key foundationalist insights and therefore can allow a key
sense in which religious beliefs enjoy direct epistemic support. Second,
because foundherentism can allow that perhaps some religious beliefs are
properly basic in unusual circumstances. Accordingly, one could plausibly
view this foundherentist proposal as merely restricting the class of religious beliefs that the parity argument has application to without thereby
discounting this manuvre altogether.
First, however, we need to get an idea of what is involved in foundherentism. In essence, it is meant to be an epistemological model that can
capture insights from both foundationalist and coherentist schools of
thought in epistemology. Consider the following passage from Haack:
Foundherentism is an intermediate theory which (unlike coherentism) allows the
relevance of experience but (unlike experientialist foundationalism) requires neither privileged beliefs justified exclusively by experience nor an essentially onedirectional notion of evidential support.34

of religious experience in mind which is much more akin to perceptual experience,

standardly understood. If this is so, then the dialectical moral to be drawn from the
modified version of reformed epistemology offered here is not that this model should
replace Plantingas own (since he is welcome to model his conception of religious belief in any way he pleases), but rather that this model generalises the core insights
within Plantingas account to make that view applicable to a broader range of religious
experience.
34
Haack 1993: 113

193
The thought is thus that foundherentism retains a core thesis from both
foundationalism and coherence theory. On the one hand, it retains the
foundationalist idea that some beliefs could enjoy sufficient positive epistemic status without that status resting upon the epistemic status of any
other beliefs (such that they are properly basic). On the other hand, it retains the coherentist idea that the positive epistemic status of some beliefs
can be the result solely of the coherence of this belief within an appropriate
set of other beliefs. The way this is achieved is by arguing that whilst some
beliefs enjoy a direct epistemic support which is not based on other beliefs, other beliefs enjoy a completely indirect transferred epistemic support which is the result of its relation to other beliefs, and still other beliefs
gain some of their epistemic support in the direct fashion and some of it in
the indirect fashion.
Haack often describes this position in terms of the metaphor of a
crossword puzzle, where the clues stand for experience. It could well be
that certain clues directly point to a certain answer (epistemically supported belief), and thus that the gaining of this answer is not dependent
upon any other answers that one might already have. This would thus represent the standard perceptual case where our perceptual beliefs directly
gain a sufficient positive epistemic status without that status being dependent upon the epistemic status of any other beliefs we hold.
In contrast, other clues wont suggest any particular answer, but one
could determine an answer by looking at the way that certain possible answers cohere with the answers already gained. A real-world example of
such a case might be a trial where the jury must form an opinion on the basis of the evidence in front of them. An epistemically supported judgement
in this case may be entirely the result of some reflective process whereby
one comes to recognise that a certain alternativethe not-guilty verdict,
saycoheres with the evidence presented in a far more adequate way than
the opposing guilty verdict.
These are two extreme cases, however, and, more usually, our belief
formation falls between these two poles. Employing the crossword metaphor again, we can say that often a clue directs us towards a small selection
of answers, and that our final determination of the correct answer is dependent upon our forming a judgement about which of these possible answers best coheres with the other answers that we already have. This thus
corresponds to cases where we form beliefs which enjoy some degree of
direct positive epistemic support, but where this support does not take one
to the threshold necessary for warrant. In such cases the epistemic support

194
in question needs to be augmented with further epistemic support from
ones other beliefs if it is to meet this threshold. An example of this might
be a case where there is some ambiguity present in experience, as when
one is unsure whether the person that one sees in the distance is ones
brother or ones father. Here, we might call on further beliefs that we hold
(that hes too tall to be ones father for example) to form our final (warranted) judgement. Foundherentism thus offers a spectrum of possibilities,
from completely direct non-inferential warrant at one extreme to completely indirect transferred warrant at the other extreme, with various degrees of combination of these two alternatives in between.35
The advantage of employing this model is that it can accommodate
the thought that religious belief enjoys some measure of direct epistemic
support (perhaps even a sufficient measure in certain rare cases), whilst
also allowing that, in general, this support is insufficient by itself to warrant the religious beliefs in question. What must be added is thus further
epistemic support from other beliefs to bring the positive epistemic status
up to the required threshold for warrant. This account thus allows us to
mark the contrast with perceptual belief (which tends to generally enjoy
sufficient direct epistemic support) in a way that accords with the intuition
that religious belief is formed in some ways that are analogous with perceptual belief.36 In particular, it allows that ones religious beliefs do enjoy
some measure of positive epistemic status that is direct and therefore not
dependent upon the epistemic status of any other beliefs that one holds. In
this sense the foundherentist modification of the reformed epistemology
thesis is just as resistant to the sceptical challenge we witnessed earlier on.
For so long as the religious belief is indeed formed in the right kind of circumstances then, even though the agent might lack good reflectively accessible grounds for his belief, it can still have, contra the sceptic, a sig35

The other advantage of this crossword metaphor is that it can offer a vivid description of belief-change, even where those beliefs were previously taken as being very
secure. As any crossword enthusiast will tell you, even the most compelling of answers to a particular clue can start to look shaky if evidence against it begins to build
in the form of several not so compelling answers that will not fit with it.
36
As DeRose has pointed out to me (in correspondence) his point is actually slightly
different from this in that his emphasis is on how religious belief is lacking in indirect
warrant relative to perceptual belief. The key point about there being a disanalogy here
between religious and perceptual belief stands either way, of course, but I retain this
particular reading of the foundherentist claim because I think that this understanding
of the difference between perceptual and religious belief best captures the disanalogies
noted earlier in 4. I am grateful to DeRose for helping me to be clear on this point.

195
nificant degree of positive epistemic status. Again, then, we find an externalist thesis emerging and it is this externalism that is doing the work of
undermining the sceptical challenge.
Furthermore, this proposal can account for why it is that certain sorts
of religious belief seem to be preferable to others. After all, if coherence in
ones religious beliefs can contribute to the epistemic status of those beliefs, then it is little wonder that an inchoate set of religious beliefs would
seem to be lacking in rationality. This approach therefore enables us to distinguish between properly held religious belief and the religious belief held
by the enthusiasts that Locke talked of.37 Evidence thus does have a central role to play in religious epistemology after all; it is just not quite as
central as the evidentialist contends. That is, the account sketched here attempts to offer a compromise view between evidentialism and the radical
anti-evidentialism espoused by Plantinga. Whereas the evidentialist sees
epistemic support as being entirely concerned with evidential considerations and Plantinga views evidential considerations as only relevant once
we have moved away from basic religious beliefs (as happens, for example, when the believer is exposed to a defeater for his basic belief), this
modified reformed account contends that evidential considerations are
nearly always relevant but usually only in concert with other epistemic
support that is direct and non-evidential. In this way evidential considerations can be accorded a role in the determination of a religious beliefs
epistemic support without this role thereby inviting the usual evidentialism-based sceptical challenge.
Finally, this account explains why it is that we rarely ascribe religious knowledge to small children, even though we are often happy to ascribe to them perceptual knowledge. The reason is that the former sort of
knowledge requires the agent to reflect on the relevant evidential considerations and therefore implicates certain reflective capacities in a way that
perceptual knowledge does not.
I think that this is an intriguing proposal, but as it stands it does not
offer us quite what we are looking for. The reason for this is that the very
schematic foundherentist account under consideration merely presents us
with an appealing description of the structural nature of the epistemology
that we want rather than going further to distinguish between the various
sorts of epistemological analyses that might fit this template and adjudicat-

37

Insofar, of course, as their beliefs really are lacking in coherence.

196
ing between them.38 Accordingly, in the next section I will try to show how
a very specific sort of epistemological positiona form of virtue epistemologycan be put into service to account for the epistemology of religious belief. This account will fit the framework offered by foundherentism whilst also being contiguous with the core formulation of reformed
epistemology due to Plantinga that is also conceived along (broadly speaking) virtue-theoretic lines.
6. A Virtue-Theoretic Proposal
The suggestion that reformed epistemology is best understood along virtue-theoretic lines may not at first seem particularly novel because, at least
in the case of Plantinga, reformed epistemology is already regarded by
some as being a form of virtue epistemology. I want to argue, however,
that reformed epistemology, where it is understood as a virtue-theoretic account, is not conceived of in terms of the right virtue-theoretic account.
Moreover, the account that I propose will fit the foundherentist template
outlined above.
What makes an epistemological account virtue-theoretic is that it is
agent-based rather than belief-based. In particular, a belief counts as
knowledge only if it is the result of, as John Greco puts it, an agents cognitively virtuous character. We can, I think, get a better handle on this distinction between belief-based and agent-based epistemology by considering some of the earliest forms of virtue-theoretic proposals in epistemology
which were expressed in terms of agent reliabilism. This kind of proposal
has been put forward by, amongst others, Ernest Sosa and Alvin Goldman.
Moreover, it is the sort of account that Plantinga himself offers and which
he explicitly applies to religious belief.39
The basic idea behind agent reliabilism is that we need to amend the
key process reliabilist accountas expressed by, for example, (an earlier)
38

Indeed, it is important to remember that the type of foundherentism that we are

dealing with here is just Haacks basic schematic account of the position rather than
the particular variant of that position which she goes on to outline. Note that this is no
criticism of DeRose since his suggestion was only meant to be structural in the first
place.
39
For the main accounts of virtue epistemology, see Sosa 1985; 1991; 1993; Montmarquet 1987; 1993; Plantinga 1988; 1993b; 1993c; 2000; Kvanvig 1992; Goldman
1993; Greco 1993; 1999; 2000; Hookway 1994; and Zagzebski 1996. See also the excellent survey article by Axtell 1997.

197
Goldman40 - along virtue-theoretic lines in order to meet some of the standard challenges to the view. For example, one of the problems that reliabilism faces is that it seems to count certain beliefs as being warranted
even when they are formed via processes which, whilst reliable, are clearly
not knowledge-conducive. There are three main examples of this ilk. For
now I will concentrate on two of them.41
The first type of counterexample to reliabilism is concerned with
those reliable processes where the success of the process does not seem to
reflect any cognitive achievement on the part of the agent. One could
imagine, for example, that an agent reliably forms true beliefs about a certain subject matter solely because some benevolent demon makes it the
case that his beliefs in this regard are reliable. In this case we have strong
intuitions that knowledge does not result because the agent is not tracking
the world in the relevant sense (instead the world appears to be tracking
the agents beliefs). And given that there is no cognitive achievement on
the part of the agent, it does not seem right to say that his reliable true beliefs can count as knowledge.
The second type of counterexample concerns certain malfunctions
on the part of the agent which, nonetheless, actually enable the agent to reliably form true beliefs about a certain subject matter.42 Because the reliability is due to a malfunction, however, we have a strong intuition that it
cannot count as being knowledge-conducive.
Of course, committed reliabilists could respond to both of these examples by modifying the view in subtle respects, but such a move would
fail to pay due attention to the heart of the difficulty here. This is that reliabilism goes wrong in only considering certain properties of the belief
rather than focussing instead upon properties of the agent who formed that
belief. In particular, the agent reliabilist thought is that not just any reliable
belief-forming process can produce knowledge, but only those processes
40

e.g., 1986.
All three of these examples are discussed, in one form or another, in Plantinga
1993a. For a general discussion of Agent Reliabilism, see Greco 1999. It is important
to note that there are a number of other advantages to adopting a virtue-theoretic epistemology that I do not have the space to expand upon here. For example, such proposals seem to be able to adequately respond to both Gettier-type scenarios and sceptical
arguments. Moreover, a virtue-theoretic epistemology may also be able to meet the socalled generality problem that has bedevilled reliabilist accounts of knowledge.
42
The example that Plantinga 1993a offers of a reliable cognitive malfunction is an
epistemically lucky brain lesion that enables the agent to form true beliefs about his
condition in this respect.
41

198
that perform certain appropriate roles within the cognitive character of the
agent. Agent reliabilists therefore argue that instead of defining knowledge
purely in terms of properties of the belief in question, one should instead
focus upon the stable natural cognitive traits, or faculties, of the agent.
Paradigm examples of such traits are our senses which, if they are working
correctly and in a stable manner relative to the appropriate environmental
conditions, will lead us to true beliefs. Plantinga, for example, characterises his version of this kind of thesis in terms of a cognitive design plan
as follows:
A belief B has warrant for S if and only if the relevant segments (the segments
involved in the production of B) are functioning properly in a cognitive environment sufficiently similar to that for which Ss faculties are designed; and the
modules of the design plan governing the production of B are (1) aimed at truth,
and (2) such that that there is a high objective probability that a belief formed in
accordance with those modules (in that sort of cognitive environment) is true;
and the more firmly S believes B the more warrant B has for S.43

What is important about such cognitive faculties is that they are more than
just (in the ideal case) reliable. In Plantingas account just cited, for example, the reliability of these faculties is merely a necessary condition for
knowledge gained through them. What is needed to make such reliability
knowledge-conducive is the further claim that this reliability arises out of a
kind of stable cognitive excellence, or virtue, that the agent exhibits and
which he can therefore take credit for.44
This conception of knowledge in terms of natural stable cognitive
faculties meets the two objections outlined above by explaining why the
agent lacks knowledge despite exhibiting a reliable belief-forming process.
In the first example, the agent cannot take any credit for his true beliefs,
and thus have beliefs which count as knowledge, because they are not due
to any trait of his, let alone a stable natural belief-forming process which
exhibits a kind of cognitive excellence. As regards the second example, the
agent reliabilist account explains why cognitive malfunctions can never
43

Plantinga 1993b: 19.

It is important to note that not everyone agrees that Plantingas account is a virtuetheoretic one, including Plantinga himself (who prefers to call his position a proper
function thesis). I think, however, that Plantingas account has enough in common
with the views expressed by the key virtue theorists to fall into this camp. For more
discussion on this point, see the exchange between Sosa 1993 and Plantinga 1993c,
and the commentary by Axtell 1997, 3-4.
44

199
give us knowledge. For not only are such processes usually lacking in the
required stability, they are also not natural cognitive faculties either. No
wonder, then, that they are unable to provide us with knowledge.
An agent reliabilist thesisa thesis which employs an understanding
of the notion of a cognitive virtue in terms of natural cognitive facultiesis thus able to meet certain objections to basic process reliabilism by
restricting the conception of what is to counts as a knowledge-producing
reliable process in ways that focus upon attributes of the agent.
Henceforth, I will refer to such early virtue epistemological accounts
as faculty virtue theories. In essence, where they differ from reliabilist accounts is merely in their stress on the agents cognitive character and their
focus on natural cognitive faculties of the agent. Such faculty virtue accounts are ideally suited to capturing perceptual knowledge because it
seems entirely uncontentious in the perceptual case to view knowledge as
being purely the result of cognitive faculties functioning correctly in the
right circumstances. Note that to conceive of perceptual knowledge in this
way is to adopt the kind of pure externalist epistemology that was discussed above. There is no demand here that the agent need bring his reflective capacities to bear in order to exhibit knowledge; instead, he need only
meet purely external conditions in order to know.
As one might expect given his use of the parity argument, Plantinga
directly employs the kind of faculty virtue account he offers for perceptual
beliefwhat he calls a proper function accountto religious belief. In
line with the Calvinist tradition to which he belongs, Plantinga argues that
we have an innate natural cognitive facultya sensus divinitatiswhich
enables us to form reliable religious beliefs and that, so long as this cognitive faculty is functioning correctly (so long as, for example, it is not adversely affected by the agents sin), then one can gain religious knowledge
in the same unmediated fashion that one gains perceptual knowledge. Perceptual beliefs and religious beliefs can thus both be thought of, at least in
the standard case, as properly basic.
We have already seen, however, that the nature of religious experience, in contrast to perceptual experience, is such that the application of a
pure externalist epistemology to religious belief is suspect. Accordingly,
Plantingas agent reliabilist account of knowledge, when applied to religious belief, will have the unfortunate consequence that an agent could
come to know certain religious propositions via the exercise of a particular
sort of cognitive faculty even though he does not bring any of his reflective
capacities to bear upon the formation of his belief in these propositions. As

200
we noted above, however, for the vast majority of religious beliefs (if not
all of them) this seems entirely unintuitive. Properly formed religious beliefs are not usually simply formed as a direct response to certain stimuli,
as one might naturally think in the case of properly formed perceptual beliefs, but instead seem to standardly invoke certain reflective capacities on
the part of the subject.45
Of course, this observation alone does not entail that the virtuetheoretic approach in general is suspect. Instead, all it shows is that those
versions of the virtue-theoretic account that cannot leave room for our reflective capacities to play an essential role in the acquisition of certain sorts
of knowledge are problematic. If there are other virtue-theoretic accounts
configured along similar lines that can allow these capacities to play the
required role, then the virtue-theoretic model will be back in business. I
think that there are such models available.
The kind of accounts that I have in mind are those virtue theories that
do not concentrate solely on the faculty virtues but also incorporate a role
for reflective virtues. Such virtues may include such cognitive traits as the
ability to weigh-up evidence impartially, or the ability to integrate ones
beliefs so as to gain, for example, a greater degree of doxastic consistency.
This general line of thought has its roots in the distinction that Sosa46
makes between brute or animalistic knowledge and reflective
knowledge, although it expands upon this basic distinction by allowing that
between these two extremes there can be various types of knowledge
which demand different combinations of brute and reflective cognitive
virtues. The thought is that by combining both faculty and reflective cognitive virtues one attains a more fine-grained account of what is involved in
knowledge possession in different cases. That is, certain sorts of knowledge, such as perceptual knowledge, might just require properly functioning cognitive faculties, whereas other sorts of knowledge, such as that
which can result from abstract reasoning for example, might solely depend
upon the reflective virtues. In between, one will find the vast majority of
knowledge that requires a mixture of both properly functioning faculty virtues and reflective virtues.47
45

For an insightful, and more general, critique of Plantingas agent reliabilism, see
Zagzebski 1996, 3.5.
46
e.g., 1991
47
One finds virtue-theoretic accounts which emphasise the importance of reflective
virtues in recent work by Greco 1993; 1999; 2000 and Zagzebski 1996, although the
stress in Grecos account is more on what he terms subjective justification than on

201
Such an account has a number of attractive features. For example,
one could use such a theory to explain why certain types of knowledge are
regarded as being more refined than others. It may be, for example, that
two knowers both meet the threshold for knowledge but that one of them
surpasses that threshold in important epistemically relevant respects, perhaps because he exhibits a certain sort of understanding of what he knows
which is lacking in the case of the other agent (and which is not simply the
result of knowing more truths than the other agent). This would thus be a
case where the refinement in question was due to the activation of a reflective capacity on the part of the subject.
A further advantage of this sort of virtue-theoretic approach is that it
can meet the third type of counterexample that is often made against process reliabilism that I alluded to earlier. This counterexample concerns
situations where one reliably gains true beliefs by forming beliefs in ways
which seem, antecedently at least, to be undesirable. For example, one
could imagine a scenario in which one forms beliefs about a certain subject
matter on the basis of bias and yet, because of some stipulated feature of
the circumstances in which the beliefs are acquired, beliefs formed on this
basis turn out to be reliable. Seemingly, however, one cannot gain knowledge through bias, no matter how reliable ones beliefs are.
This example should sound familiar because it is a variation of the
example offered earlier in our critique of evidentialism. There it was noted
that evidentialism (in at least some of its forms at any rate) is unable to
make any adequate distinction between those agents who believe what they
epistemically ought to believe for all the wrong reasons (such as because of
bias), and those who believe what they epistemically ought to believe for
all the right reasons (because of a sensitivity to the weight and extent of
their evidence, for example). Again, then, we have a situation in which an
agent has met the relevant epistemic rubric that has been set, but has nevertheless met it in a way that seems to preclude that agent from possessing
knowledge.
Agent reliabilism lacks the resources to respond to examples of this
sort because there does not seem to be any way in which one can trace the
reflective virtues as such. Moreover, although the accounts offered by Greco and Zagzebski are in this respect similar to that sketched here, they do tend to go further to
make the activation of such reflective capacities necessary for knowledge. In contrast,
my claim is much weaker in that I allow that in certain casessuch as in the perceptual casean agent might have knowledge without exhibiting any reflective capacities
at all.

202
cognitive shortcoming in question back to the agents cognitive faculties
where these are understood in non-reflective terms. The more developed
form of virtue epistemology under consideration here has no such difficulties, however, because it can explain the agents lack of knowledge in
terms of a failure to exhibit the appropriate reflective virtue. In this case,
for example, the agent forms his beliefs in terms of bias (a reflective vice)
rather than in response to the weight and extent of his evidence (as reflective virtue would dictate). The same goes for the evidentialist variant of
this example. If one characterises knowledge in terms of true belief that
arises out of a cognitive virtue, then those agents who believe what they
epistemically ought to believe for all the wrong reasons lack knowledge
precisely because that true belief (if it is true) does not arise out of a reflective virtue.
The attraction of applying such a thesis to religious knowledge
should be clear. For example, a reflective virtue epistemology of this type
can allow that whilst it might be true in the perceptual case that one can
gain knowledge without exhibiting any developed reflective capacities, the
same need not also be true in the case of religious knowledge. This view
can thus do justice to our intuition that, at least in the standard case, the
role of the faculty virtues alone as regards religious belief is insufficient to
afford us religious knowledge. Instead, a precondition of acquiring religious knowledge (at least in the standard case) will be that the agent has
(and brings to bear) the appropriate reflective virtues as well. An account
of this sort is therefore able to capture the reformed intuition that religious
knowledge is similar to perceptual knowledge in certain respects, in that
both forms of knowledge presuppose that the agent has certain properly
functioning faculty virtues, whilst also allowing that this analogy is not
complete and thus that some epistemological explanation of the disanalogies present here should also be given. These disanalogies are accounted
for in terms of the role played by the reflective virtues in the acquisition of
religious knowledge.48
7. Concluding Remarks
In effect, what we have done here is show how knowledge in general can
48

Indeed, it may be that we need to incorporate not only reflective virtues into this
account but also moral ones as well, in that religious knowledge is the sort of knowledge which might directly implicate certain virtuous moral traits. For more on this
point, see Zagzebski 1996.

203
be adequately understood in terms of a virtue-theoretic account that employs both faculty and reflective virtues, and then further shown that such
an account can accommodate the nature of religious belief, especially as
regards the respects in which it differs from perceptual belief. Moreover,
since we are retaining the key externalist thought that knowledge need not
be dependent upon the agent being able to adduce sufficient non-questionbegging evidence, we are no longer troubled by an evidentialism-based
sceptical argument concerning religious belief. It is still true that the legitimacy of an agents religious belief is not solely dependent upon that
agent being able to adduce evidence appropriate and proportionate to the
belief in question. All that is different about this account as opposed to
Plantingas properly basic account is that if an agents religious belief is
to enjoy a warrant then that agent must exhibit the relevant reflective virtues as well, and this will typically involve the ability to adduce some appropriate degree of evidence. Far from being a sceptical hurdle for religious belief to clear, however, this condition can actually serve to distinguish responsible religious belief from the undisciplined religious belief of
Lockes enthusiast. Provided that this reflective condition is met then,
ultimately, whether or not an agents belief is actually warranted will depend upon whether the appropriate external facts obtain, just as in the
perceptual case.
Note also that this suggestion is entirely in the spirit of the foundherentist proposal made by DeRose, since a natural way of modelling the different types of knowledge here, whether faculty-virtue-based, reflectivevirtue-based, or a mixture of the two, is in terms of a foundherentist structure. What we have done is merely add specific content to the basic foundherentist structure by offering a particular epistemological proposal that is
structured along these lines that is both independently plausible and which
can accommodate important features of religious experience and the belief
that it gives rise to. Moreover, since the virtue-theoretic proposal offered
here is merely an extension of the sort of earlier virtue-theoretic account
offered by Plantinga, it ought to be in the spirit of the general reformed
epistemological approach. Adapting our understanding of the parity argument thus gives rise to a reformed conception of reformed epistemology
that is able to meet at least one of the key problems that the unreformed
version faces.49
49

An earlier version of this paper was presented at The Epistemology of Basic Belief conference, Vrije Universiteit, Amsterdam, The Netherlands, in June 2001.
Thanks to the audience that day and, in particular, to Andrew McGonigal, Christian

204
REFERENCES
Alston, W. P. 1982. Religious Experience and Religious Belief, Nos 16: 3-12.
. 1986. Is Religious Belief Rational? in The Life of Religion, edited by
S. M. Harrison and R. C. Taylor, 1-15. Lanham, Maryland: University Press of
America.
.1991. Perceiving God: The Epistemology of Religious Experience. Cornell,
Ithaca: Cornell University Press.
Axtell, G. 1997. Recent Work on Virtue Epistemology, American Philosophical
Quarterly 34: 1-26.
BonJour, L. 1985. The Structure of Empirical Knowledge. Cambridge,
Massachusetts: Harvard University Press.
Chisholm, R. 1989. Theory of Knowledge, third edition. Englewood Cliffs,
New Jersey: Prentice Hall.
Clifford, W. K. 1879. Lectures and Essays, edited by F. Pollock. London: Macmillan.
DeRose, K. 1999a. Are Christian Beliefs Properly Basic? Unpublished manuscript.
.1999b. Voodoo Epistemology. Unpublished manuscript.
. 2000. Ought We to Follow Our Evidence? Philosophy and Phenomenological Research 60: 697-706.
Feldman, R. and Conee, E. 1985. Evidentialism, Philosophical Studies 48: 15-34.
Feldman, R. 2000. The Ethics of Belief, Philosophy and Phenomenological Research 60: 667-96.
Goldman, A. 1986. Epistemology and Cognition. Cambridge, Massachusetts:
Harvard University Press.
. 1993. Epistemic Folkways and Scientific Epistemology, Philosophical IsMiller, Alvin Plantinga, and to the chair, Hunter Davies. Thanks also to Michael
Brady, Keith DeRose and an anonymous referee for this volume who read and commented on an earlier draft of this paper.

205
sues 3:271-84.
Greco, J. 1993. Virtues and Vices of Virtue Epistemology, Canadian Journal of
Philosophy 23: 413-32.
. 1999. Agent Reliabilism, Philosophical Perspectives 13: 273-96.
.2000. Putting Skeptics in Their Place: The Nature of Skeptical Arguments and
Their Role in Philosophical Inquiry. Cambridge, UK: Cambridge University
Press.
Haack, S. 1993. Double-Aspect Foundherentism: A New Theory of Empirical Justification, Philosophy and Phenomenological Research 53: 113-28.
. 1995. Evidence and Inquiry: Towards Reconstruction in Epistemology. Oxford: Basil Blackwell.
Hookway, C. 1994. Cognitive Virtues and Epistemic Evaluations, International
Journal of Philosophical Studies 2: 211-27.
Kornblith, H. 2001. Epistemology: Internalism and Externalism (ed.). Oxford: Basil
Blackwell.
Kvanvig, J. 1992. The Intellectual Virtues and the Life of the Mind. Savage,
Maryland: Rowman & Littlefield.
Locke, J. 1979. Essay Concerning Human Understanding, edited by P. H. Nidditch.
Oxford: Oxford University Press.
McLeod, M. S. 1993. Rationality and Theistic Belief: An Essay on Reformed
Epistemology. Ithaca, New Jersey: Cornell University Press.
Miller, C. B. 2004. Defeaters and the Basicality of Theistic Belief. This volume.
Mitchell, B. 1973. The Justification of Religious Belief. Philadelphia: Temple
University Press.
Montmarquet, J. 1987. Epistemic Virtue, Mind 96: 487-97.
. 1993. Epistemic Virtue and Doxastic Responsibility. Lanham, Maryland:
Rowman & Littlefield.
Newman, J. H. 1844. Sermons, Chiefly on the Theory of Religious Belief, Preached
Before the University of Oxfor. London.

206
. 1985. An Essay in Aid of a Grammar of Assent, edited by I. T. Ker. Oxford:
Oxford University Press.
Nielsen, K. 1967. Wittgensteinian Fideism, Philosophy 42: 237-54.
Penelhum, T. 1983. God and Skepticism. Dordrecht, Netherlands: Reidel.
. 1986. Do Religious Beliefs Need Grounds?, Nederlands Theologisch
Tijdschrift 40: 227-37.
. 2000. Reflections on Reformed Epistemology. Unpublished manuscript.
Phillips, D. Z. 1976. Religion Without Explanation. Oxford: Oxford University Press.
Plantinga, A. 1980. The Reformed Objection to Natural Theology, Proceedings of
the American Catholic Philosophical Association 49-62. New York.
.1981. Is Belief in God Properly Basic?, Nos 15: 41-51.
.1983. Reason and Belief in God in Faith and Rationality, edited by
A. Plantinga and N. Wolterstorff, 16-93. Notre Dame, Indiana: University of
Notre Dame.
.1988. Positive Epistemic Status and Proper Function, Philosophical
Perspectives 2: 1-50.
. 1993a. Warrant: The Current Debate. New York: Oxford University Press.
. 1993b. Warrant and Proper Function. New York: Oxford University Press.
. 1993c. Why We Need Proper Function, Nos 27: 66-82.
. 2000. Warranted Christian Belief. New York: Oxford University Press.
Plantinga, A. and Wolterstorff, N. 1983. Faith and Rationality, editors, Notre Dame,
Indiana: University of Notre Dame.
Pritchard, D. H. 2000. Is God Exists a Hinge Proposition of Religious Belief?
International Journal for Philosophy of Religion 47: 129-40.
. 2002a. McKinsey Paradoxes, Radical Scepticism, and the Transmission of
Knowledge across Known Entailments, Synthese 130: 279-302.
. 2002b. Recent Work on Radical Skepticism, American Philosophical
Quarterly 40: 215-57.

207
Sosa, E. 1985. Knowledge and Intellectual Virtue, The Monist 68: 224-45.
.1991. Intellectual Virtue in Perspective in Knowledge in Perspective: Selected Essays in Epistemology, E. Sosa. Cambridge, UK: Cambridge University
Press.
.1993. Proper Functionalism and Virtue Epistemology, Nos 27: 51-65.
Swinburne, R. 1979. The Existence of God. Oxford: Oxford University Press.
. 1996. Is There a God?. Oxford: Oxford University Press.
Wittgenstein, L. 1966. Wittgensteins Lectures and Conversations on Aesthetics,
Psychology and Religious Belief, edited by C. Barrett. Oxford: Basil Blackwell.
Wolterstorff, N. 1991. The Migration of the Theistic Arguments: From Natural
Theology to Evidentialist Apologetics. In Contemporary Classics in
Philosophy of Religion edited by A. Loades and L. D. Rue, 71-92. La Salle,
Illinois: Open Court.
. 1995. Divine Discourse: Philosophical Reflections on the Claim that God
Speaks. Cambridge, UK: Cambridge University Press.
.1996. John Locke and the Ethics of Belief. Cambridge, UK: Cambridge
University Press.
Wright, C. 1985. Facts and Certainty, Proceedings of the British Academy 71:
429-72.
. 2000. Cogency and Question-Begging: Some Reflections on McKinseys
Paradox and Putnams Proof, Philosophical Issues 10: 140-63.
Zagzebski, L. 1996. Virtues of the Mind: An Inquiry into the Nature of Virtue and the
Ethical Foundations of Knowledge. Cambridge, UK: University of Cambridge
Press.